Design Of Real-time Recognition System For Android Data Stream Based On Decision Tree Algorithm

The development of information technology and the wide use of smart phones provide great convenience for people's work and life.With the enrichment of mobile phone applications,network traffic has become complex and diversified.How to optimize the specific application scenarios of Android smart phones to meet the needs of users is a problem worthy of in-depth study.The premise of network optimization is to classify and identify the data traffic in the network.The traditional research on data flow identification mainly focuses on the communication protocol or application type in the application layer,and does not consider the fine-grained classification,and the recognition efficiency has not been tested.Most of the research is carried out on the application layer,and the identification efficiency has not been tested.In view of the above problems,this paper makes an in-depth study on the data stream identification of Android mobile applications.This paper mainly does two parts of work.In the first part,the basic definition of data flow and the evaluation standard of real-time identification are given,and then the data of popular apps are collected on the PC side.According to the typical functions provided by the application,the data flow of each app is divided into two or three more refined traffic types,and the identification model of APP traffic is built based on C4.5 decision tree algorithm.The data sets of each application are processed from the aspects of data cleaning,stream feature extraction,model training,etc.,and the model is evaluated from the accuracy of recognition.Experimental results show that when the number of data stream features is 4 or 5,the decision tree algorithm can keep the recognition accuracy of specific traffic types in each application above 99%,and the algorithm also has strong real-time performance.In the second part,based on the decision tree model trained in the first part,data stream recognition function is realized in Android mobile kernel.Because Android system is developed based on Linux kernel,this paper uses Netfilter framework provided by Linux to capture and process packets,and the information of the data packet is extracted through the SK?BUFF structure under the kernel which stores and manages the data packet.The decision tree model of each application is implemented in the kernel data flow identification code according to IF-THEN rule,and the extracted data flow information is input into the model for identification.The actual test is carried out on Google mobile phone,running the program,loading the module through the ADB debugging tool,and printing the program running log.The results show that the data flow identification system implemented in the kernel can accurately and quickly identify the running application traffic types,and the accurate identification time is less than 200 ms,which meets the real-time requirements of data flow identification.In addition,the data stream identification function is compiled into kernel module,which is loaded and unloaded dynamically in the form of module,which improves the expansibility and maintainability of the system function.In general,the Android data stream recognition system based on C4.5 decision tree algorithm achieves good results in fine-grained traffic recognition,and has strong real-time performance in Android mobile kernel.The recognition model has the characteristics of high accuracy,light weight and easy expansion,and is suitable for real Android smart phone scenarios.